signalHacker News Show HN2026-09-23
Show HN: Castrag - transcribe and semantically search a large podcast archive
Castrag is a resumable, open-source pipeline (MIT) for transcribing and querying large audio archives via the OpenAI API, processing 230 episodes (~145 hours) for about $33. It produces per-episode structured summaries, a semantic search index (numpy array + JSON, no vector DB), and RAG-based Q&A with grounded citations using scripts like transcribe.py, summarize.py, build_index.py, ask.py, and analyze_themes.py. The workflow uses gpt-4o-mini-transcribe, chunks audio to ~15-minute segments, and allows resumable, concurrent execution.
- for who
- Podcasters, researchers, or anyone with large audio archives who wants searchable and queryable content.
- why now
- No clear timeliness signal - article lacks a time-sensitive trigger.
- what changes
- Large untranscribed audio archives become instantly searchable and queryable without manual listening or expensive full-context LLM calls.
- to do
- Clone the repo, configure OpenAI API key, drop audio files into audio-files/, and run the provided scripts in sequence.
key points
- Processed 230 episodes (~145h) for $33 via OpenAI API
- Offers structured summaries, semantic search, and cited RAG Q&A
- Uses numpy array index, no vector database, MIT license
#podcast transcription#semantic search#rag#open source tool#cost optimization
score
score 8 out of 10. 0-10: how dense the facts are, multiplied by how much you can do with them after reading. 8+ means the topic's evidence bar is met: benchmarks and availability for a new model, amount and investors for a funding round, revenue figures for a solo-money story. Below 5 an item does not enter the digest. A press release scores 3 or less, a reprint loses 2, anything older than 14 days loses 1, a headline that misleads loses 3.
read the source